bl1_dotv2axpyv2b.c File Reference

(r)

Functions
void	bl1_sdotv2axpyv2b (int n, float *a1, int inc_a1, float *a2, int inc_a2, float *x, int inc_x, float *kappa1, float *kappa2, float *rho1, float *rho2, float *w, int inc_w)
void	bl1_ddotv2axpyv2b (int n, double *a1, int inc_a1, double *a2, int inc_a2, double *x, int inc_x, double *kappa1, double *kappa2, double *rho1, double *rho2, double *w, int inc_w)
	if (inc_a1 !=1||inc_a2 !=1||inc_x !=1||inc_w !=1)
	for (i=0;i< n_run;++i)
	if (n_left > 0)
void	bl1_cdotv2axpyv2b (int n, scomplex *a1, int inc_a1, scomplex *a2, int inc_a2, scomplex *x, int inc_x, scomplex *kappa1, scomplex *kappa2, scomplex *rho1, scomplex *rho2, scomplex *w, int inc_w)
void	bl1_zdotv2axpyv2b (int n, dcomplex *a1, int inc_a1, dcomplex *a2, int inc_a2, dcomplex *x, int inc_x, dcomplex *kappa1, dcomplex *kappa2, dcomplex *rho1, dcomplex *rho2, dcomplex *w, int inc_w)

Variables
double *restrict	alpha2
double *restrict	chi1 = x
double *restrict	omega1 = w
double	kappa1_c = *kappa1
double	kappa2_c = *kappa2
double	rho1_c
double	rho2_c
int	i
int	n_pre
int	n_run
int	n_left
*	rho1 = rho1_c
*	rho2 = rho2_c
	alpha1 = a1
rho1_c	real = 0.0
rho1_c	imag = 0.0

Function Documentation

bl1_cdotv2axpyv2b()

void bl1_cdotv2axpyv2b	(	int	n,
		scomplex *	a1,
		int	inc_a1,
		scomplex *	a2,
		int	inc_a2,
		scomplex *	x,
		int	inc_x,
		scomplex *	kappa1,
		scomplex *	kappa2,
		scomplex *	rho1,
		scomplex *	rho2,
		scomplex *	w,
		int	inc_w
	)

{
    bl1_abort();
}

References bl1_abort().

bl1_ddotv2axpyv2b()

void bl1_ddotv2axpyv2b	(	int	n,
		double *	a1,
		int	inc_a1,
		double *	a2,
		int	inc_a2,
		double *	x,
		int	inc_x,
		double *	kappa1,
		double *	kappa2,
		double *	rho1,
		double *	rho2,
		double *	w,
		int	inc_w
	)

{
    double*   restrict alpha1;
    double*   restrict alpha2;
    double*   restrict chi1;
    double*   restrict omega1;
    double             rho1_c;
    double             rho2_c;
    int                i;
 
    int                n_pre;
    int                n_run;
    int                n_left;
    
    v2df_t    k1v, rho1v;
    v2df_t    k2v, rho2v;
    v2df_t    a11v, a12v, x1v, w1v;
    v2df_t    a21v, a22v, x2v, w2v;
    
    if ( inc_a1 != 1 ||
         inc_a2 != 1 ||
         inc_x  != 1 ||
         inc_w  != 1 ) bl1_abort();
 
    n_pre = 0;
    if ( ( unsigned long ) a1 % 16 != 0 )
    {
        if ( ( unsigned long ) a2 % 16 == 0 ||
             ( unsigned long ) x % 16 == 0 ||
             ( unsigned long ) w % 16 == 0 ) bl1_abort();
 
        n_pre = 1;
    }
 
    n_run       = ( n - n_pre ) / 4;
    n_left      = ( n - n_pre ) % 4;
 
    alpha1   = a1;
    alpha2   = a2;
    chi1     = x;
    omega1   = w;
 
    rho1_c = 0.0;
    rho2_c = 0.0;
 
    if ( n_pre == 1 )
    {
        double   kappa1_c = *kappa1;
        double   kappa2_c = *kappa2;
        double   alpha1_c   = *alpha1;
        double   alpha2_c   = *alpha2;
        double   chi1_c     = *chi1;
        double   omega1_c   = *omega1;
 
        rho1_c   += alpha1_c * chi1_c;
        omega1_c += kappa1_c * alpha1_c;
 
        rho2_c   += alpha2_c * chi1_c;
        omega1_c += kappa2_c * alpha2_c;
 
        *omega1 = omega1_c;
 
        alpha1   += inc_a1;
        alpha2   += inc_a2;
        chi1     += inc_x;
        omega1   += inc_w;
    }
 
    rho1v.v = _mm_setzero_pd();
    rho2v.v = _mm_setzero_pd();
 
    k1v.v = _mm_loaddup_pd( ( double* )kappa1 );
    k2v.v = _mm_loaddup_pd( ( double* )kappa2 );
 
    for ( i = 0; i < n_run; ++i )
    {
        a11v.v = _mm_load_pd( ( double* )alpha1 );
        a12v.v = _mm_load_pd( ( double* )alpha2 );
        x1v.v  = _mm_load_pd( ( double* )chi1 );
        w1v.v  = _mm_load_pd( ( double* )omega1 );
 
        rho1v.v += a11v.v * x1v.v;
        w1v.v += k1v.v * a11v.v;
 
        rho2v.v += a12v.v * x1v.v;
        w1v.v += k2v.v * a12v.v;
 
        _mm_store_pd( ( double* )omega1, w1v.v );
 
        a21v.v = _mm_load_pd( ( double* )(alpha1 + 2) );
        a22v.v = _mm_load_pd( ( double* )(alpha2 + 2) );
        x2v.v  = _mm_load_pd( ( double* )(chi1 + 2) );
        w2v.v  = _mm_load_pd( ( double* )(omega1 + 2) );
 
        rho1v.v += a21v.v * x2v.v;
        w2v.v += k1v.v * a21v.v;
 
        rho2v.v += a22v.v * x2v.v;
        w2v.v += k2v.v * a22v.v;
 
        _mm_store_pd( ( double* )(omega1 + 2), w2v.v );
 
        alpha1   += 4;
        alpha2   += 4;
        chi1     += 4;
        omega1   += 4;
    }
 
    if ( n_left > 0 )
    {
        for ( i = 0; i < n_left; ++i )
        {
            double   kappa1_c = *kappa1;
            double   kappa2_c = *kappa2;
            double   alpha1_c   = *alpha1;
            double   alpha2_c   = *alpha2;
            double   chi1_c     = *chi1;
            double   omega1_c   = *omega1;
 
            rho1_c   += alpha1_c * chi1_c;
            omega1_c += kappa1_c * alpha1_c;
 
            rho2_c   += alpha2_c * chi1_c;
            omega1_c += kappa2_c * alpha2_c;
 
            *omega1 = omega1_c;
 
            alpha1   += inc_a1;
            alpha2   += inc_a2;
            chi1     += inc_x;
            omega1   += inc_w;
        }
    }
 
    rho1_c += rho1v.d[0] + rho1v.d[1];
    rho2_c += rho2v.d[0] + rho2v.d[1];
 
    *rho1 = rho1_c;
    *rho2 = rho2_c;
}

References alpha1, alpha1_c, alpha2, alpha2_c, bl1_abort(), chi1, v2df_t::d, i, kappa1_c, kappa2_c, n_left, n_pre, n_run, omega1, rho1, rho1_c, rho2, rho2_c, and v2df_t::v.

Referenced by FLA_Fused_Ahx_Ax_opd_var1(), and FLA_Fused_UZhu_ZUhu_opd_var1().

bl1_sdotv2axpyv2b()

void bl1_sdotv2axpyv2b	(	int	n,
		float *	a1,
		int	inc_a1,
		float *	a2,
		int	inc_a2,
		float *	x,
		int	inc_x,
		float *	kappa1,
		float *	kappa2,
		float *	rho1,
		float *	rho2,
		float *	w,
		int	inc_w
	)

{
    bl1_abort();
}

References bl1_abort().

bl1_zdotv2axpyv2b()

void bl1_zdotv2axpyv2b	(	int	n,
		dcomplex *	a1,
		int	inc_a1,
		dcomplex *	a2,
		int	inc_a2,
		dcomplex *	x,
		int	inc_x,
		dcomplex *	kappa1,
		dcomplex *	kappa2,
		dcomplex *	rho1,
		dcomplex *	rho2,
		dcomplex *	w,
		int	inc_w
	)

{
    dcomplex* restrict alpha1;
    dcomplex* restrict alpha2;
    dcomplex* restrict chi1;
    dcomplex* restrict omega1;
    int                i;
 
    v2df_t    kappa1v, kappa1rv;
    v2df_t    kappa2v, kappa2rv;
    v2df_t    rho1v;
    v2df_t    rho2v;
    v2df_t    a11v, a12v;
    v2df_t    a21v, a22v;
    v2df_t    x1v, x1rv;
    v2df_t    w1v;
    v2df_t    acbc, bdad;
    v2df_t    adac, bcbd;
 
    if ( inc_a1 != 1 ||
         inc_a2 != 1 ||
         inc_x  != 1 ||
         inc_w  != 1 ) bl1_abort();
 
    alpha1   = a1;
    alpha2   = a2;
    chi1     = x;
    omega1   = w;
 
    rho1v.v = _mm_setzero_pd();
    rho2v.v = _mm_setzero_pd();
 
    kappa1v.v  = _mm_load_pd( ( double* )kappa1 );
    kappa1rv.v = _mm_shuffle_pd( kappa1v.v, kappa1v.v, _MM_SHUFFLE2 (0,1) );
    kappa2v.v  = _mm_load_pd( ( double* )kappa2 );
    kappa2rv.v = _mm_shuffle_pd( kappa2v.v, kappa2v.v, _MM_SHUFFLE2 (0,1) );
 
    for ( i = 0; i < n; ++i )
    {
        //dcomplex omega1_c = *omega1;
        w1v.v  = _mm_load_pd( ( double* )omega1 );
 
        //dcomplex chi1_c   = *chi1;
        x1v.v  = _mm_load_pd( ( double* )chi1 );
 
 
        //dcomplex alpha1_c = *alpha1;
        a11v.v  = _mm_loaddup_pd( ( double* )&(alpha1->real) );
        a12v.v  = _mm_loaddup_pd( ( double* )&(alpha1->imag) );
 
        //rho1_c.real += alpha1_c.real * chi1_c.real - -alpha1_c.imag * chi1_c.imag;
        //rho1_c.imag += alpha1_c.real * chi1_c.imag + -alpha1_c.imag * chi1_c.real;
        x1rv.v = _mm_shuffle_pd( x1v.v, x1v.v, _MM_SHUFFLE2 (0,1) );
        adac.v = a11v.v * x1rv.v;
        bcbd.v = a12v.v * x1v.v;
        rho1v.v = rho1v.v + _mm_addsub_pd( adac.v, bcbd.v );
 
        //omega1_c.real += kappa1_c.real * alpha1_c.real - kappa1_c.imag * alpha1_c.imag;
        //omega1_c.imag += kappa1_c.real * alpha1_c.imag + kappa1_c.imag * alpha1_c.real;
        acbc.v = kappa1v.v  * a11v.v;
        bdad.v = kappa1rv.v * a12v.v;
        w1v.v += _mm_addsub_pd( acbc.v, bdad.v );
 
 
        //dcomplex alpha2_c = *alpha2;
        a21v.v  = _mm_loaddup_pd( ( double* )&(alpha2->real) );
        a22v.v  = _mm_loaddup_pd( ( double* )&(alpha2->imag) );
 
        //rho2_c.real += alpha2_c.real * chi1_c.real - -alpha2_c.imag * chi1_c.imag;
        //rho2_c.imag += alpha2_c.real * chi1_c.imag + -alpha2_c.imag * chi1_c.real;
        x1rv.v = _mm_shuffle_pd( x1v.v, x1v.v, _MM_SHUFFLE2 (0,1) );
        adac.v = a21v.v * x1rv.v;
        bcbd.v = a22v.v * x1v.v;
        rho2v.v = rho2v.v + _mm_addsub_pd( adac.v, bcbd.v );
 
        //omega1_c.real += kappa2_c.real * alpha2_c.real - kappa2_c.imag * alpha2_c.imag;
        //omega1_c.imag += kappa2_c.real * alpha2_c.imag + kappa2_c.imag * alpha2_c.real;
        acbc.v = kappa2v.v  * a21v.v;
        bdad.v = kappa2rv.v * a22v.v;
        w1v.v += _mm_addsub_pd( acbc.v, bdad.v );
 
 
        //*omega1 = omega1_c;
        _mm_store_pd( ( double* )omega1, w1v.v );
 
 
        //alpha1   += inc_a1;
        //alpha2   += inc_a2;
        //chi1     += inc_x;
        //omega1   += inc_w;
        alpha1   += 1;
        alpha2   += 1;
        chi1     += 1;
        omega1   += 1;
    }
 
    rho1v.v = _mm_shuffle_pd( rho1v.v, rho1v.v, _MM_SHUFFLE2 (0,1) );
    rho2v.v = _mm_shuffle_pd( rho2v.v, rho2v.v, _MM_SHUFFLE2 (0,1) );
 
    //*rho1 = rho1_c;
    //*rho2 = rho2_c;
    _mm_store_pd( ( double* )rho1, rho1v.v );
    _mm_store_pd( ( double* )rho2, rho2v.v );
}

References alpha1, alpha2, bl1_abort(), chi1, i, omega1, rho1, rho2, and v2df_t::v.

Referenced by FLA_Fused_Ahx_Ax_opz_var1().

for()

for

(

)

    {
        double   alpha11_c   = *alpha1;
        double   alpha21_c   = *(alpha1 + 1);
        double   alpha12_c   = *alpha2;
        double   alpha22_c   = *(alpha2 + 1);
        double   chi1_c     = *chi1;
        double   chi2_c     = *(chi1 + 1);
        double   omega1_c   = *omega1;
        double   omega2_c   = *(omega1 + 1);
 
        // rho1 += conj(alpha1) * chi1;
        rho1_c += alpha11_c * chi1_c;
        rho1_c += alpha21_c * chi2_c;
 
        // omega1 += kappa1 * alpha1;
        omega1_c += kappa1_c * alpha11_c;
        omega2_c += kappa1_c * alpha21_c;
 
        // rho2 += conj(alpha2) * chi1;
        rho2_c += alpha12_c * chi1_c;
        rho2_c += alpha22_c * chi2_c;
 
        // omega1 += kappa2 * alpha2;
        omega1_c += kappa2_c * alpha12_c;
        omega2_c += kappa2_c * alpha22_c;
 
        *omega1       = omega1_c;
        *(omega1 + 1) = omega2_c;
 
        alpha1   += 2*inc_a1;
        alpha2   += 2*inc_a2;
        chi1     += 2*inc_x;
        omega1   += 2*inc_w;
    }

References alpha1, alpha2, chi1, kappa1_c, kappa2_c, omega1, rho1_c, and rho2_c.

if() [1/2]

if

(

inc_a1 !

= 1 || inc_a2 != 1 || inc_x != 1 || inc_w != 1

)

    {
        double   alpha1_c   = *alpha1;
        double   alpha2_c   = *alpha2;
        double   chi1_c     = *chi1;
        double   omega1_c   = *omega1;
 
        rho1_c   += alpha1_c * chi1_c;
        omega1_c += kappa1_c * alpha1_c;
 
        rho2_c   += alpha2_c * chi1_c;
        omega1_c += kappa2_c * alpha2_c;
 
        *omega1 = omega1_c;
 
        alpha1   += inc_a1;
        alpha2   += inc_a2;
        chi1     += inc_x;
        omega1   += inc_w;
    }

if() [2/2]

if	(	n_left	,
		0
	)

    {
        for ( i = 0; i < n_left; ++i )
        {
            double   alpha1_c   = *alpha1;
            double   alpha2_c   = *alpha2;
            double   chi1_c     = *chi1;
            double   omega1_c   = *omega1;
 
            rho1_c   += alpha1_c * chi1_c;
            omega1_c += kappa1_c * alpha1_c;
 
            rho2_c   += alpha2_c * chi1_c;
            omega1_c += kappa2_c * alpha2_c;
 
            *omega1 = omega1_c;
 
            alpha1   += inc_a1;
            alpha2   += inc_a2;
            chi1     += inc_x;
            omega1   += inc_w;
        }
    }

References alpha1, alpha1_c, alpha2, alpha2_c, chi1, i, kappa1_c, kappa2_c, n_left, omega1, rho1_c, and rho2_c.

Variable Documentation

alpha1

alpha1 = a1

Referenced by bl1_ddotv2axpyv2b(), bl1_zdotv2axpyv2b(), for(), and if().

alpha2

dcomplex *restrict alpha2

Initial value:

{
    double*   restrict alpha1

Referenced by bl1_daxpyv2b(), bl1_daxpyv3b(), bl1_ddotv2axpyv2b(), bl1_zaxpyv2b(), bl1_zdotv2axpyv2b(), FLA_Fill_with_geometric_dist(), FLA_Fill_with_inverse_dist(), FLA_Fill_with_logarithmic_dist(), FLA_Fused_Ahx_Axpy_Ax_opd_var1(), FLA_Fused_Ahx_Axpy_Ax_opz_var1(), FLA_Fused_Gerc2_Ahx_Axpy_Ax_opd_var1(), FLA_Fused_Gerc2_Ahx_Axpy_Ax_opz_var1(), for(), and if().

chi1

chi1 = x

Referenced by bl1_ddotv2axpyv2b(), bl1_zdotv2axpyv2b(), for(), and if().

i

int i

Referenced by bl1_ddotv2axpyv2b(), bl1_zdotv2axpyv2b(), and if().

imag

rho2_c imag = 0.0

kappa1_c

dcomplex kappa1_c = *kappa1

Referenced by bl1_ddotv2axpyv2b(), for(), and if().

kappa2_c

dcomplex kappa2_c = *kappa2

Referenced by bl1_ddotv2axpyv2b(), for(), and if().

n_left

int n_left

Referenced by bl1_ddotv2axpyv2b(), and if().

n_pre

int n_pre

Referenced by bl1_ddotv2axpyv2b().

n_run

int n_run

Referenced by bl1_ddotv2axpyv2b().

omega1

omega1 = w

Referenced by bl1_ddotv2axpyv2b(), bl1_zdotv2axpyv2b(), for(), and if().

real

rho2_c real = 0.0

rho1

* rho1 = rho1_c

Referenced by bl1_ddotv2axpyv2b(), and bl1_zdotv2axpyv2b().

rho1_c

dcomplex rho1_c

Referenced by bl1_ddotv2axpyv2b(), for(), and if().

rho2

* rho2 = rho2_c

Referenced by bl1_ddotv2axpyv2b(), and bl1_zdotv2axpyv2b().

rho2_c

dcomplex rho2_c

Referenced by bl1_ddotv2axpyv2b(), for(), and if().